Detecting changes in multiple sclerosis (MS) lesions through follow-up MR images is an important but time-consuming and subjective process. In this work, we propose a fully automatic deep learning-based method to detect new MS lesions. The model was trained and tuned using the MSSEG2 challenge dataset. First, the brain and spinal cord masks were generated, and registration between two time points was performed. Then, new lesions and whole lesions were segmented by patch-wise inputs, respectively. The final mask for new lesions was produced by comparing these two segmentations, and in this way we could effectively reduce false-positives.
This abstract and the presentation materials are available to members only; a login is required.